Aug 2026· Frontiers in Oncology· Vol 16· 0 citations· 44 references
Medicine
TL;DR
The experimental results demonstrate that the optimized DCE-MRI–based radiomics model effectively predicts breast cancer molecular subtypes and the gradient boosting decision tree (GBDT) model combined with rigorous feature selection shows high predictive performance, highlighting its strong potential for noninvasive, accurate, and clinically applicable molecular classification of breast cancer.
Abstract
Breast cancer is a heterogeneous disease, and accurate preoperative identification of molecular subtypes is essential for guiding individualized treatment and prognostic evaluation. However, histopathological assessment, the current gold standard, is invasive and limited by intratumoral heterogeneity, underscoring the need for reliable noninvasive alternatives. Radiomics based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has shown promise for molecular subtype prediction; nevertheless, the optimal radiomic features for automated molecular classification remain to be fully elucidated. To address this gap, this study aims to identify optimal DCE-MRI radiomic features and to develop radiomics-based machine learning models for predicting breast cancer molecular subtypes. Specifically, regions of interest were manually delineated on DCE-MRI images for feature extraction, followed by feature selection to identify the most informative radiomic parameters. Four machine learning classifiers (RF, SVM, LR, GBDT) were then constructed to predict five molecular subtypes of breast cancer. The experimental results demonstrate that the optimized DCE-MRI–based radiomics model effectively predicts breast cancer molecular subtypes. In particular, the gradient boosting decision tree (GBDT) model combined with rigorous feature selection shows high predictive performance, highlighting its strong potential for noninvasive, accurate, and clinically applicable molecular classification of breast cancer.
Background The preoperative differentiation of lung adenocarcinoma subtypes is critical for implementing personalized treatment but is difficult to accomplish with conventional imaging. This study aimed to develop an interpretable multimodal model integrating clinical, peritumoral, radiomic, and deep learning features to improve diagnostic accuracy. Methods A total of 3,038 patients from four hospitals were divided into training (n=1,822), test (n=608), and validation (n=608) sets. Two radiologists manually segmented two-dimensional tumor regions on computed tomography using ITK-SNAP software. After Pearson correlation analysis and least absolute shrinkage and selection operator regression, the radiomic score and deep learning score were generated. Clinical features were selected via univariate analysis, the Boruta algorithm, and recursive feature elimination (RFE). Individual logistic models were built and fused with the optimal combination selected via support vector machine-synthetic minority oversampling technique and extreme gradient boosting. Performance was evaluated in terms of the Obuchowski index, accuracy, F1-score, calibration, and decision curves, while interpretability was assessed via Shapley additive explanations (SHAP) and individual conditional expectation (ICE). Results The fused model achieved Obuchowski indices of 0.85 [95% confidence interval (CI): 0.84–0.87], 0.81 (95% CI: 0.78–0.83), and 0.79 (95% CI: 0.76–0.81) in the training, test, and validation sets, respectively outperforming the single-modality models. The F1-scores for the lepidic, acinar/papillary, and solid/micropapillary subtypes, respectively, were 0.77, 0.61, and 0.63 in the training set; 0.72, 0.57, and 0.59 in the test set; and 0.74, 0.54, and 0.52 in the validation set. Calibration and decision curve analysis confirmed the robustness and clinical utility of the model. SHAP analysis identified ResNet-101 feature as the best predictor, followed by peritumoral radiomic score, and lobulation. ICE plots revealed the linear and monotonic relationships between key features and predicted probabilities across subtypes. Conclusions The radiomics model developed in this study facilitates the accurate and interpretable preoperative classification of lung adenocarcinoma subtypes. Fusion of clinical, peritumoral, and deep learning features enhances diagnostic performance and supports clinical decision-making.
Feng-Juan Tian, Jing Ding, Zhen-Yu Cao et al.· Quantitative Imaging in Medi...· 0 citations
Introduction Reliable differentiation between malignant and benign prostate lesions remains a critical challenge in clinical practice, particularly in settings with limited access to expert interpretation of multiparametric MRI (mpMRI). We investigated whether a streamlined radiomics framework based solely on apparent diffusion coefficient (ADC) imaging could provide accurate, reproducible, and interpretable diagnostic support for prostate cancer detection. Methods We retrospectively analyzed ADC images from 198 men, including 121 patients with histologically confirmed prostate cancer and 77 patients without malignancy, acquired between 2020 and 2025. Following image resampling and standardization, 107 radiomic features were extracted and subjected to hierarchical feature selection using recursive feature elimination followed by LASSO regression. Four discriminative features were retained to train support vector machine, random forest, logistic regression, and gradient boosting decision tree (GBDT) classifiers. Model performance was evaluated in independent training and testing cohorts using accuracy, AUC, precision, recall, F1 score, and specificity. Results The GBDT model demonstrated the most robust performance, achieving AUC of 0.81 (95% CI: 0.75–0.87) in the training cohort and 0.78 (95% CI: 0.72–0.85) in the testing cohort. In the testing cohort, the model achieved an accuracy of 0.76 (95% CI: 0.69–0.84), a recall of 1.00 (95% CI: 1.00–1.00), an F1 score of 0.86 (95% CI: 0.80–0.92), and a specificity of 0.73 (95% CI: 0.62–0.88). The selected radiomic features reflected image energy, textural irregularity, and intralesional heterogeneity and were consistently associated with malignant pathology. Discussion A minimal-feature radiomics model derived exclusively from ADC images showed promising diagnostic performance for differentiating malignant from benign prostate lesions. By avoiding reliance on multiparametric imaging, this streamlined approach may provide a practical, reproducible, and resource-efficient strategy for supporting prostate cancer diagnosis, particularly in settings where expert mpMRI interpretation is limited. Further multicenter validation, integration of clinical variables, and incorporation of explainable artificial intelligence approaches are warranted to establish its robustness and enhance its translational potential.
Kun Zhang, Jia-Jun Zhang, Yangguang Yuan et al.· Frontiers in Oncology· 0 citations
Objective Accurate identification of complete response (CR) after neoadjuvant chemoradiotherapy (NCRT) is essential for selecting candidates for watch-and-wait treatment in advanced rectal cancer. This study aimed to develop and evaluate a magnetic resonance imaging (MRI) radiomics-based machine learning model to classify CR and Non-CR in post-NCRT rectal MRI images. Methods Using region-of-interest masks, 107 radiomic features were extracted and normalized to a 0–1 range using min-max scaling. Four feature selection methods (ANOVA, RFE, SBS, and LASSO) were paired with four classifiers (LR, SVM, RF, and XGB), and all combinations were evaluated through 5-fold cross-validation in the training set using mean ROC AUC as the comparison metric. Results Among 16 combinations, RFE + SVM achieved the highest cross-validation AUC of 0.89. On an independent test set, the model achieved a sensitivity of 0.84, specificity of 0.80, accuracy of 0.82, and F1-score of 0.82. The most informative features were intensity and local texture features, including TotalEnergy, 10Percentile, Coarseness, Strength, Correlation, ZoneEntropy, and Idmn. Conclusions MRI radiomics combined with machine learning shows preliminary promise as an exploratory approach for classifying CR and Non-CR after NCRT in rectal cancer. Given the retrospective, single-center design and the absence of external validation, these findings should be interpreted as preliminary, and prospective multicenter validation is required before clinical implementation. A multicenter external validation integrating endoscopy and digital rectal examination findings is planned to enhance generalizability and clinical utility.
Jin-Young Min, Jun Young Park, Young Jae Kim et al.· Digital Health· 0 citations
Accurate assessment of Ki-67 expression levels in breast cancer is crucial for determining prognosis and making informed treatment decisions. Current immunohistochemical methods relying on needle biopsy introduce sampling errors due to tumor spatial heterogeneity, making the development of non-invasive, precise preoperative prediction methods of significant clinical importance. This study aims to explore and compare advanced deep learning models based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) for noninvasive assessment of Ki-67 expression. This retrospective study analyzed preoperative DCE-MRI data from 308 patients with histologically confirmed breast cancer. Adjacent slices centered on the tumor’s most significant cross-section were obtained to create a 2.5-dimensional (2·5D) dataset. We innovatively developed two deep learning models using the same dataset (1): a Multi-Instance Learning (MIL) model that combines slice-level predictive features with Predictive Likelihood Histogram (PLH) and Bag-of-Words (BoW) techniques (2); a Transformer-based fusion model that directly captures global contextual relationships between slices via self-attention mechanisms. The predictive performance of both models was systematically compared with traditional radiomics and clinical models. On an independent test set, the Transformer fusion model demonstrated optimal predictive performance with an area under the curve (AUC) of 0.875, achieving accuracy, sensitivity, and specificity of 0.839, 0.848, and 0.833, respectively. The MIL model ranked second (AUC = 0.825), with both models significantly outperforming traditional radiomics models (AUC = 0.698) and clinical models (AUC = 0.648). Deep learning models based on 2·5D DCE-MRI, especially Transformer models that achieve global feature fusion through self-attention mechanisms, can effectively and non-invasively predict Ki-67 expression status in breast cancer, surpassing traditional methods. This model shows potential as a reliable tool to help clinicians accurately assess tumor proliferation activity before surgery.
Yiying Cao, Mi Lin, Yanshan Ouyang et al.· Cancer Imaging· 0 citations
Background Multiparametric magnetic resonance imaging (mpMRI) is an established component of prostate cancer diagnosis; however, PI-RADS interpretation remains partly subjective, particularly for equivocal lesions. Radiomics can quantify imaging characteristics that may not be readily appreciated visually, while explainable artificial intelligence (XAI) can make machine-learning predictions more transparent. Methods A retrospective cohort of 300 men undergoing prostate mpMRI followed by histopathological assessment was hypothetically evaluated. T2-weighted (T2W), apparent diffusion coefficient (ADC), and dynamic contrast-enhanced (DCE) sequences were analyzed according to PI-RADS v2.1. Shape, first-order, and texture radiomic features were extracted following standardized preprocessing. Reproducible features were selected using intraclass correlation, correlation filtering, and LASSO. Logistic regression, random forest, support vector machine, and XGBoost models were assessed. SHAP was used to explain feature contributions. Results In the illustrative dataset, 132/300 patients (44.0%) had clinically significant prostate cancer (csPCa; Grade Group ≥2). The integrated PI-RADS, clinical, and radiomics model achieved an illustrative AUC of 0.95 (95% CI, 0.92–0.98), sensitivity of 90.5%, specificity of 88.1%, and accuracy of 89.3% in the test cohort. ADC entropy, T2W texture heterogeneity, ADC percentile measures, DCE texture features, and lesion morphology were among the leading predictors. Conclusions An explainable mpMRI radiomics framework may provide complementary quantitative information to PI-RADS and clinical variables for prediction of csPCa. SHAP-based explanations can link model predictions to anatomically plausible imaging characteristics. Prospective multicenter external validation is required before clinical implementation.
Sacheena Kumbara, Sudha Y. Soudi, Pramod Kumar V· EPRA international journal o...· 0 citations
RATIONALE AND OBJECTIVES
Breast cancer exhibits high biological heterogeneity and variable invasion patterns. Tumor budding (TB) is a key histopathological marker of aggressive behavior and poor prognosis. However, preoperative TB assessment is limited by biopsy sampling issues. This study aims to develop a multiparametric magnetic Resonance Imaging (MRI)-based habitat radiomics model for noninvasive preoperative prediction of TB status, providing an imaging biomarker to support personalized surgical planning.
MATERIALS AND METHODS
This retrospective study included 281 breast cancer patients who underwent preoperative multiparametric MRI. Patients were divided into a training set (n = 175), an internal validation set (n = 76), and an external validation set (n = 30). Intratumoral habitats were generated by combining voxel-based K-means clustering with a 3 × 3 × 3 neighborhood expansion, resulting in 16 subregions per patient, from which radiomic features were extracted. Multiple TB prediction models were constructed using two machine learning algorithms and evaluated comprehensively. Furthermore, a nomogram was built by integrating key clinical factors and heterogeneity features to assess clinical utility, and the model was subjected to SHapley Additive exPlanations (SHAP) interpretability analysis.
RESULTS
Multivariate analysis identified Ki-67 and enhancement pattern as independent predictors of TB. The combined model (logistic regression [LR]) achieved optimal performance, with area under the receiver operating characteristic curve values of 0.927, 0.858, and 0.858 in the training, internal validation, and external validation sets, respectively. SHAP analysis revealed that T2 habitat elongation, diffusion-weighted imaging habitat flatness, enhancement pattern, and Ki-67 were the core contributing features of the model. Higher elongation, lower flatness, and higher Ki-67 levels were associated with an increased risk of TB, providing potentially interpretable biological evidence for model decision-making.
CONCLUSION
An interpretable, multiparametric MRI-based intratumoral habitat heterogeneity machine learning model can noninvasively predict TB status preoperatively, offering a reliable reference for individualized treatment planning in breast cancer patients.
Min Sun, Weining Zhao, Yuwei Wang et al.· Academic Radiology· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
